Nuclear magnetic resonance has become a key medical tool for clinical diagnosis, monitoring, and intervention. This foremost imaging modality is routinely used to assess a broad range of pathologies which include breast cancer, glioblastoma brain tumours, neurodegenerative diseases, and knee lesions. In contrast to other imaging modalities such as computed tomography (CT) or X-ray, magnetic resonance does not utilise ionising radiation. Thereby, avoiding harm to patients. Its non-invasive character and excellent soft tissue contrast allow detecting diseases earlier and with higher precision than ever.However, access to magnetic resonance is restricted to a large extent because, to increase sensitivity, systems tend to employ high magnetic fields. Strong fields create compatibility conflicts with metallic implants and other medical instrumentation. More importantly, increasing the static magnetic field makes this technology more expensive and difficult to site due to higher upfront and maintenance costs, and increased safety concerns. Its use is therefore limited to advanced hospitals, making the technology beyond the reach of many patients throughout the world. Even major hospitals have constrained magnetic resonance resources, forcing them to prioritise their usage and exploiting only a fraction of benefits afforded by magnetic resonance technology.Ultra-low field magnetic resonance promises to be a more cost-effective alternative to conventional magnetic resonance systems as its hardware is simpler. Moreover, it is a good candidate for a mobile solution due to its smaller size, lower power consumption, lower weight, and reduced safety concerns. It is also more compatible with other instruments and can assess patients with implanted or lodged metals. Moreover, its frequency of operation provides unique resonance conditions which can open up novel applications to elucidate chemical or biological processes, such as directly mapping neuronal activity.The main components of conventional ultra-low field magnetic resonance systems are a shielding box that reduces interference with foreign magnetic fields, resistive coils to generate a range of magnetic fields, detectors to sense the magnetic resonance signal, and a console that governs the system. Generated magnetic fields are a strong pre-polarisation field to increase the intensity of the signal received, an adjustable measurement field which defines acquisition conditions, a radio frequency field to induce the signal, and a spatially varying linear gradient field which encodes the signal in space to produce images. Yet, although the power required to generate these fields is lower in an ultra-low field system than in a high field counterpart, the energy requirement of ultra-low field systems is still significant and require pricey power amplifiers, which constrain systems' portability and increase their costs. Besides, ultra-low field systems intrinsically suffer from a smaller signal, which has motivated the use of sophisticated sensors such as superconducting quantum interference devices or atomic magnetometers. Notably, superconducting quantum interference devices require cryogenics, and atomic magnetometers have to be carefully isolated from strong pulsating magnetic fields. These sensing technologies are fragile, bulky, and expensive, hindering system portability, versatility, and affordability.This PhD project aims to develop methods leading towards a ‘truly' portable and low-cost ultra-low field magnetic resonance instrument. Current work focuses on the generation of the magnetic fields and the signal detection as these two major parts set the lower boundary for achievable signal quality, contrast versatility, and system dimensions, and delimit the requirements of the remaining parts.First, instead of using resistive coils to generate magnetic fields, we propose employing permanent magnets which can be dynamically repositioned. We arrange the permanent magnets forming a cylindrical configuration which can generate an adjustable measurement field and a switchable strong pre-polarisation field. This field versatility has been validated empirically with a static prototype. We also study the possibility of generating gradient fields through additional permanent magnets. Results suggest that it is possible to generate a 3D magnetic resonance image with a minimal number of magnets moved along simple trajectories.The second hardware development is comprised of methods to design inexpensive and robust detectors with exquisite sensitivity and rapid activation. High sensitivities are achieved by optimising the detectors with more accurate numerical models, exploring a more extensive range of arrangements, and considering most significant elements. These detectors traditionally delay acquisitions in the presence of the pulsed magnetic fields employed in magnetic resonance experiments. Here, this drawback is significantly attenuated through the combination of hardware and software solutions contributing to higher acquisition efficiency. A prototype resulting from these methods has empirically been tested, which validates the proposed models.Proposed magnets design provides required quasi-static magnetic fields in a compact and energy efficient manner without compromising system safety or decreasing field versatility. In addition, the detector methods developed provide perhaps the highest acquisition efficiency value amongst ultra-low field magnetic resonance sensors, allowing for room temperature imaging with inherently robust detectors. Here presented novel methods pave the way towards a clinically relevant, low-cost, and portable ultra-low field magnetic resonance system.